jimmyg1997 / corona_tscs

This is the data repository of the CoronaNet project on government responses to the COVID-19 pandemic and the data/code repository for the paper "A Retrospective Bayesian Model for Measuring Covariate Effects on Observed COVID-19 Test and Case Counts".

README

CoronaNet Project Team April 24th, 2020

About This Repository

This repository contains data from the CoronaNet data collection project and also data and code to fit the model described in “A Retrospective Bayesian Model for Measuring Covariate Effects on Observed COVID-19 Test and Case Counts”, link here. Following is first a list of data for the CoronaNet project, with data dictionary, and subsequently a list of files relevant to “A Retrospective Bayesian Model for Measuring Covariate Effects on Observed COVID-19 Test and Case Counts”.

On the CoronaNet Update Tracker, you can track our policy updates by country and subnational unit.

The following plot shows our policy activity index, a set of scores produced by a dynamic measurement model from our data. It is also included in the data release and is a helpful way to reduce the data to a single score. It also permits more straightforward inter-country comparisons.

CoronaNet Data

First, CoronaNet data releases:

Please note that while we make every effort to validate this data, the speed and scale with which it was collected means that we cannot validate all of it. If you find an error in the data, please file an issue on this Github page.

The format of the data is in country-day-record_id format. Some record_id values have letters appended to indicate that the general policy category type also has a value for type_sub_cat, which contains more detail about the policy, such as whether health resources refers to masks, ventilators, or hospitals. Some entries are marked as new_entry in the entry_type field for when a policy of that type was first implemented in the country. Later updates to those policies are marked as updates in entry_type. To see how policies are connected, look at the policy_id field for all policies from the first entry through updates for a given country/province/city. If an entry was corrected after initial data collection, it will read corrected in the entry_type field (the original incorrect data has already been replaced with the corrected data).

1. data/CoronaNet/data_bulk/coronanet_release[.rds/csv.gz] These files contain variables from the CoronaNet government response project, representing national and sub-national policy event data from more than 140 countries since January 1st, 2020. The data include source links, descriptions, targets (i.e. other countries), the type and level of enforcement, and a comprehensive set of policy types. For more detail on this data, you can see our codebook here.

2. data/CoronaNet/data_bulk/coronanet_release_allvars[.rds/csv.gz] These files contains the government response information from coronanet_release.csv along with the following datasets:

   1. Tests from the CoronaNet testing database (see http://coronanet-project.org for more info);
   2. Cases/deaths/recovered from the JHU data repository (https://github.com/CSSEGISandData/COVID-19);
   3. Country-level covariates including GDP, V-DEM democracy scores, human rights indices, power-sharing indices, and press freedom indices from the Niehaus World Economics and Politics Dataverse (https://niehaus.princeton.edu/news/world-economics-and-politics-dataverse)

3. data/CoronaNet/data_country/coronanet_release_[country].csv For each country in coronanet_release, we have generated a separate data file in a .csv format.

4. data/CoronaNet/data_country/coronanet_release_allvars_[country].csv For each country in coronanet_release_allvars, we have generated a separate data file in a .csv format.

coronanet_release.csv Field Dictionary

1. record_id Unique identifier for each unique policy record
2. policy_id Identifier linking new policies with subsequent updates to policies
3. recorded_date When the record was entered into our data
4. date_updated When we can confirm the country - policy type was last checked/updated (we can only confirm policy type for a given country is up to date as of this date)
5. date_announced When the policy is announced
6. date_start When the policy goes into effect
7. date_end When the policy ends (if it has an explicit end date)
8. entry_type Whether the record is new, meaning no restriction had been in place before, or an update (restriction was in place but changed). Corrections are corrections to previous entries.
9. event_description A short description of the policy change
10. domestic_policy Indicates where policy targets an area within the initiating country (i.e. is domestic in nature)
11. type The category of the policy
12. type_sub_cat The sub-category of the policy (if one exists)
13. type_text Any additional information about the policy type (such as the number of ventilators/days of quarantine/etc.)
14. index_high_est The high (95% posterior density) estimate of the country policy activity score (0-100)
15. index_med_est The median (most likely) estimate of the country policy activity score (0-100)
16. index_low_est The low (95% posterior density) estimate of the country policy activity score (0-100)
17. index_country_rank The relative rank by each day for each country on the policy activity score
18. country The country initiating the policy
19. init_country_level Whether the policy came from the national level or a sub-national unit
20. province Name of sub-national unit
21. target_country Which foreign country a policy is targeted at (i.e. travel policies)
22. target_geog_level Whether the target of the policy is a country as a whole or a sub-national unit of that country
23. target_region The name of a regional grouping (like ASEAN) that is a target of the policy (if any)
24. target_province The name of a province targeted by the policy (if any)
25. target_city The name of a city targeted by the policy (if any)
26. target_other Any geographical entity that does not fit into the targeted categories mentioned above
27. target_who_what Who the policy is targeted at
28. target_direction Whether a travel-related policy affects people coming in (Inbound) or leaving (Outbound)
29. travel_mechanism If a travel policy, what kind of transportation it affects
30. compliance Whether the policy is voluntary or mandatory
31. enforcer What unit in the country is responsible for enforcement
32. link A link to at least one source for the policy
33. ISO_A3 3-digit ISO country codes
34. ISO_A2 2-digit ISO country codes

coronanet_release_allvars.csv Field Dictionary

1. All of the fields listed above, plus

2. tests_daily_or_total Whether a country reports the daily count of tests a cumulative total

3. tests_raw The number of reported tests collected from host country websites or media reports

4. deaths The number of COVID-19 deaths, aggregated to the country-day level (JHU CSSE data)

5. confirmed_cases The number of confirmed cases of COVID-19, aggregated to the country-day level (JHU CSSE data)

6. recovered The number of recoveries from COVID-19, aggregated to the country-day level (JHU CSSE data)

7. ccode The Correlates of War country code

8. ifs IMF IFS country code

9. Rank_FP (most recent year available from Niehaus dataset) Reporters without Borders Press Freedom Annual Ranking

10. Score_FP (most recent year available from Niehaus dataset) Reporters with Borders Press Freedom Score

11. state_IDC (most recent year available from Niehaus dataset) State/Provincial Governments Locally Elected

12. muni_IDC (most recent year available from Niehaus dataset) Municipal Governments Locally Elected

13. dispersive_IDC (most recent year available from Niehaus dataset) Dispersive Powersharing

14. constraining_IDC (most recent year available from Niehaus dataset) Constraining Powersharing

15. inclusive_IDC (most recent year available from Niehaus dataset) Inclusive powersharing

16. sfi_SFI (most recent year available from Niehaus dataset) State fragility index

17. ti_cpi_TI (most recent year available from Niehaus dataset) Corruption perceptions index

18. pop_WDI_PW (most recent year available from Niehaus dataset) World Bank population

19. gdp_WDI_PW (most recent year available from Niehaus dataset) World Bank GDP (total)

20. gdppc_WDI_PW (most recent year available from Niehaus dataset) World Bank GDP per capita

21. growth_WDI_PW (most recent year available from Niehaus dataset) World Bank GDP growth percent

22. lnpop_WDI_PW (most recent year available from Niehaus dataset) Log of World Bank population

23. lngdp_WDI_PW (most recent year available from Niehaus dataset) Log of World Bank GDP

24. lngdppc_WDI_PW (most recent year available from Niehaus dataset) Log of World Bank GDP per capita

25. disap_FA (most recent year available from Niehaus dataset) 3 category, ordered variable for disappearances index

26. polpris_FA (most recent year available from Niehaus dataset) 3 category, ordered variable for political imprisonment index

27. latentmean_FA (most recent year available from Niehaus dataset) the posterior mean of the latent variable index for human rights protection)

28. transparencyindex_HR (most recent year available from Niehaus dataset) Transparency Index

29. EmigrantStock_EMS (most recent year available from Niehaus dataset) Total emmigrant stock from

30. v2x_polyarchy_VDEM (most recent year available from Niehaus dataset) Electoral democracy index

31. news_WB (most recent year available from Niehaus dataset) Daily newspapers (per 1,000 people)

A Retrospective Bayesian Model for Measuring Covariate Effects Data and Code

Files to reproduce the paper:

1. retrospective_model_paper/corona_tscs_betab.stan: This Stan model contains a partially-identified model of COVID-19 that permits relative distinctions to be made between areas/countries/states’ infection rates. The parameter num_infected_high indexes the infection rate by time point and country. As the latent process is on the logit scale, it must be converted via the inverse-logit function to a proportion. However, the resulting estimate should not be interpreted as the total infected in a country, but rather a relative ranking of which countries/areas are the most infected up to the current time point.

2. retrospective_model_paper/corona_tscs_betab_scale.stan: This Stan model extends the partially-identified model with the 10% lower threshold for tests to infections ratio described in the paper. This model will produce an estimate for num_infected_high that when converted with the inverse-logit function will represent the proportion infected in a country conditional on the model’s prior concerning the tests to infections ratio.

3. retrospective_model_paper/kubinec_model_preprint.Rmd: A copy of the paper draft with embedded R code. You can access fitted Stan model objects to compile the paper here: https://drive.google.com/open?id=1cTCQTAjH8I-11jp3CEdIJZ0NaGRAn8dT. Otherwise all Stan models must be re-fit to compile the paper. The process will take approximately 2 hours.

4. retrospective_model_paper/kubinec_model_SI.Rmd: This file contains an Rmarkdown file with embedded R code showing how to simulate the model. It is the supplementary information for the paper. See the compiled .pdf version as well.

5. data: The data folder contains CSVs of tests and cases for US states and other data that were used to fit the models in the paper.

6. retrospective_model_paper/BibTexDatabase.bib: This file contains the Bibtex bibliography for the paper.